EP0898298A1

EP0898298A1 - Determination of the thickness of a denuded zone in a silicon wafer

Info

Publication number: EP0898298A1
Application number: EP97830354A
Authority: EP
Inventors: Maria Luisa Polignano; Marzio Brambilla; Francesco Cazzaniga; Giuseppe Pavia; Federica Zanderigo
Original assignee: STMicroelectronics SRL; SGS Thomson Microelectronics SRL
Current assignee: STMicroelectronics SRL
Priority date: 1997-07-15
Filing date: 1997-07-15
Publication date: 1999-02-24
Anticipated expiration: 2017-07-15
Also published as: US6197606B1; DE69738129D1; EP0898298B1

Abstract

The depth of a denuded layer in respect to a relatively defective bulk region of a monocrystalline semiconductor wafer is estimated in a nondestructive way by measuring effective diffusion length of injected excess minority charge carriers on a surface of the wafer having such a denuded layer and on a different portion of the surface of the wafer from where the denuded layer has been previously stripped-off by lapping and/or etching and by calculating the depth of the denuded layer through a best-fit procedure or through numerical processing of the measurement results on the basis of the diffusion equations of excess minority carriers.

Description

Field of Application of the Invention

The present invention relates to fabrication processes of integrated circuits and more in particular to methods for controlling the quality of monocrystalline silicon substrates.

Background of the Invention

The silicon generally used for fabricating integrated circuits is grown using a technique referred to as the Czochralski (CZ) technique and therefore contains oxygen in a concentration comprised generally between 5 and 10*10¹⁷ atoms cm^-3 (ASTM 83 units) .

A substantially oxygen-free silicon monocrystal may be grown by other techniques such as the so-called "flow-zone" technique but an oxygen-free silicon is unsuitable for fabricating integrated circuits because the presence of oxygen improves its mechanical properties.

On the other hand, the presence of oxygen may generate defects in the crystalline structure of silicon unless great care is exercised in exposing the substrate to particularly critical conditions during the fabricating process of integrated circuits.

Indeed, in CZ silicon oxygen may result oversaturated at particularly elevated process temperatures and precipitate during thermal treatments.

Oxygen precipitation induces the formation of extensive defects (dislocations, staking faults) and may degrade the performance of integrated circuits if these defects occur in the active areas of the devices. By contrast, if the defects induced by oxygen precipitation grow sufficiently far from the active areas of the devices, for instance deep in the bulk of the monocrystal wafer, they tend to have a positive effect as they act as gathering centers of metal impurities (atoms of transition metals of the periodic table) . For these reasons, treatment techniques have been developed for obtaining silicon wafers where defects induced by oxygen precipitation exist only in an innermost region (bulk) in respect to the wafer thickness, and have oxygen defect-free superficial layers (denuded zone) .

Wafers so processed are then used for fabricating integrated devices essentially in the defect-free surface region also referred to as "denuded-zone" of one of the two sides of the wafer of monocrystalline silicon.

The thickness of the denuded zone depends on the properties of the starting material, on the denuding pre-treatment and/or in many cases on the thermal treatments that are contemplated in the specific fabrication process of the integrated devices, so that it becomes necessary to monitor the actual thickness of the denuded zone to ensure that the active regions be reliably defined in defect-free regions on the silicon wafer.

Moreover, it has been recognized that oxygen precipitation in the bulk should be moderate in order to be beneficial . Thence, the density and size of oxygen precipitate in the bulk of a wafer should be relatively small and the techniques nowadays used to measure the thickness of denuded zones often fail under conditions of insufficient density of oxygen precipitate density in the bulk.

Denuded zone thicknesses are commonly measured by means of microscopy techniques.

The monocrystal cleavage, the selective etching and inspections of the samples by a suitable microscopy technique (optical microscopy, scanning electron microscopy (SEM), atomic force microscopy or others) are limited by an insufficient selectivity of the etching of the sample surface.

Transmission Electron Microscopy (TEM) does not require a selective etching of the samples and possesses a very high sensitivity, though it is statistically limited. Indeed, TEM may provide for a relatively correct estimate of the depth of the denuded zone only when the defect density in the bulk is sufficiently large, because of the restricted region of a sample that may be explored with this technique.

As mentioned above, the density and size of defects in the bulk of the wafer of monocrystalline silicon tend, for other reasons, to be particularly small in present days wafers and as a consequence the monitoring of the depth of the denuded zone by the known techniques becomes problematic.

Moreover, these methods are destructive and require proper sample preparation which is a burdensome operation, especially for TEM exploration. For these reasons, reliably producing depth maps of denuded zones is at present difficult and costly.

In theory, the monitoring of the denuded zone thickness could be performed, by electrical methods, under favorable conditions. Probably, the most reliable of these techniques is the so-called Electron Beam Induced Current (EBIC). This method is very sensitive but time consuming, besides it requires a laborious sample preparation. In any case, it remains a destructive technique. Therefore, the EBIC technique remains unsuitable as effective monitoring technique and for producing maps of the depth of a denuded zone in terms of compatibility of the time required and of costs to implement a reliable process quality control.

On the other hand, it is well known that lifetime measurements of current carriers in a semiconducting silicon monocrystal are very sensitive to the presence of defects in the crystal lattice of the semiconductor and hence to the presence of oxygen precipitates in the silicon, especially if the lifetime measurements are carried out under conditions of low injection of electric charges.

According to a technique and relative measuring equipment, known in the trade with the name of Elymat , excess minority carriers are generated by a laser beam that illuminates the front of a wafer and are collected in the space charge region of a Schottky contact which may be established either on the wafer back-side (backside photocurrent or BPC) or on the same illuminated or frontside of the wafer(frontside photocurrent of FPC) .

The sample is dipped in a diluted HF solution to establish a Schottky contact on one side of the wafer passivate the surface layer states on the other side. The injection level δn/p_o (where δn is the concentration of excess minority carriers and p_o is the equilibrium concentration of minority carriers) can be varied by varying the power of the illuminating laser. Carrier lifetime (τ) or diffusion length (Ldiff = Dτ , where D is the minority carriers diffusivity) data are extracted from photocurrent measurements. FPC measurements are used when the lifetime in the sample under consideration is so low that practically no current could be gathered at a contact established on the backside of the wafer. This is often the case in presence of oxygen precipitates, and thus FPC measurements are more suited than BPC measurements in view of the objective of the present invention.

A particularly sensitive method for measuring the lifetime under low injection conditions is the so-called method of Surface Photovoltage Measurement (SPV).

Surface photovoltage measurements are carried out by illuminating the sample with lights of different wavelengths. The minority carriers generated are collected in a depletion region on the same wafer surface (at a certain distance from the illuminated area) and produce a variation of the surface potential, which is recorded in function of the wavelength of the illuminating light and therefore in function of the penetration depth of the radiation in the semiconductor crystal.

At present there exist methods of surface photovoltage measurement that are particularly fast and do not require sample preparation.

These methods of lifetime measurement by surface photovoltage measurements are intrinsically rapid and do not require costly sample preparation and therefore have been indicated as potentially suitable, likewise other techniques, to provide more or less reliable and approximate estimates of the depth of a defect-free superficial layer where it exists.

The method of surface photovoltage measurements, that is of carrier lifetime measurements, has been proposed as a way to assess the depth of a defect-free superficial layer of a wafer of semiconducting monocrystalline silicon when the density of the defects induced by oxygen precipitates in the bulk of the silicon is sufficiently high .

This technique has been found unsuitable to provide for reliable estimates of the depth of the denuded zone in the case of wafers generally used at present that are characterized by having a relatively low density of oxygen precipitates in their bulk region.

Scope and Summary of the Invention

Faced with these difficulties and limitations of known methods, has now been found a simple and effective nondestructive method for determining the depth of a denuded zone of a wafer of semiconducting monocrystal that produces reliable measurements of the depth of a denuded zone even in the presence of relatively small densities and sizes of oxygen precipitates in the bulk, that would make inoperable the electrical methods known at present.

Essentially, the nondestructive method of the invention consists in:

stripping off by mechanical lapping and/or by wet or dry etching the denuded layer of the semiconductor wafer from at least a portion of the surface, preferably from the backside (where it can be removed from the whole surface) or from a certain area of the frontside of the wafer;
measuring a surface photovoltage signal in function of the injection depth of charge carriers "injected" by a photoelectric effect or by an equivalent injection method in a certain area of the wafer surface, on a portion of surface having such denuded layer and on a portion stripped of said denuded layer;
solving through a nonlinear "best-fit" procedure based on the measured values of the surface photovoltage signal function and using as a fitting parameter the thickness of the denuded zone or layer, or through equivalent numerical methods of the diffusion equation of charge carriers injected in the two distinct portions of surface.

Although any technique for measuring the effective diffusion length or lifetime of charge carriers, suitable for low injection level may be satisfactorily utilized, preferably the lifetime measurements of excess minority carriers injected in a defined semiconductor zone, are performed by surface photovoltage methods.

While a "best-fit" technique with an analytical function obtained by solving the charge carrier diffusion equation, is a relatively simple procedure, it is obvious that modern numerical processing will yield comparable or even better results.

Brief Description of the Drawings

Figure 1, as already mentioned, shows a typical map and relevant characteristics and values of the diffusion length of minority carriers as normally obtained with an Elymat apparatus.

Figure 2 shows the dependence on the level of injection of the mean value of minority carriers lifetime.

Figure 3 shows the influence of oxygen precipitates on measured minority carrier lifetime.

Figure 4 shows the distribution profiles of excess minority charge carriers in function of depth in a wafer with a denuded zone, obtained using superficial photovoltage measurements.

Figure 5 shows the surprising coincidence with the real pattern of the SPV signal in function of the penetration depth of the incident light and the data best-fit with the function calculated on the base of the diffusion equation for a region including both the bulk and the denuded zone.

Description of the Method of the Invention and Comparison with Different Measuring Techniques and Relative Results

These tests were carried out using a commercial apparatus for surface photovoltage measurements manufactured by the company SDI model CSM III-A SPV. This instrument is equipped with a set of filters that produce monochromatic lights of seven different wavelengths, ranging from 800nm to 1mm. The corresponding light penetration depth in a silicon crystal range approximately between 11µm and 157µm.

Since the intrinsic lifetime measurements may only give indirect estimates of the depth of an eventual denuded zone of a wafer, that is of a substantially defect-free superficial layer of the wafer, the results obtained with the method of the invention were compared with direct observations carried out on the same samples by conventional destructive methods, based on microscopic observation techniques, in order to have a measure of the degree of reliability of the indirectly data obtained by the method of the invention. Hence, after performing lifetime measurements on sample wafers, the wafers were cut, cleaved, selectively etched with an etchant solution known in the trade as "Secco d'Aragona" and finally inspected by a Scanning Electron Microscope (SEM) or by Atomic Force Microscope (AFM). The same samples were also inspected by Transmission Electron Microscopy (TEM).

According to an essential aspect of the method of the invention, the denuded zone was removed completely from the backside of the sample wafers by lapping, and the resulting damage caused by lapping was eliminated by etching the silicon for removing the highly stressed layer left by the lapping process.

The total thickness stripped off was about 150µm which is surely greater than the expected depth of a denuded zone of a wafer treated for this purpose.

After having stripped off the denuded zone from the backside of the silicon wafer, lifetime measurements were carried out for comparison purposes via photocurrent measurements on the backside of sample wafers, using an FPC Elymat equipment. The resulting diffusion length of minority carriers was found to be much smaller than the thickness of the bulk region of the wafer containing oxygen precipitate defects so that this diffusion length provided a measure of the recombination efficiency in a defective region of silicon.

Fig. 1 shows a typical map of minority carriers diffusion lengths obtained by an FPC Elymat technique in the bulk region of the wafer silicon containing oxygen precipitates. The concentric ring aspect of the map reflecting the crystal growth conditions is immediately recognized by a person conversant with the Elymat technique and instrumentation and represents a "marker" of oxygen precipitation.

The measurements of the charge carriers lifetime were analyzed in function of the injection level for some of the most common contaminants of the silicon in order to identify for certainty the dominant recombination center.

Fig. 2 shows the measured lifetime which differed by about one order of magnitude among the samples, however the same dependence of the lifetime on the injection level δn/p_o was verified in all the samples.

This kind of behavior was always observed in association with a map of concentric rings as the one shown in Fig. 1, therefore the dominant recombination center could be identified in oxygen precipitation sites.

These tests demonstrate that the recombination efficiency of oxygen precipitates is higher the lower is the level of injection.

A similar lifetime dependence on the injection level was also observed in virgin CZ silicon wafers, that is in wafers that were not treated to produce denuded zones thereon, coming from three different suppliers. In silicon monocrystals obtained with the CZ method, nuclei of oxygen precipitates tend to form during ingot cooling and during the heat treatment for disactivating thermal donors so that it is not unlikely that the oxygen precipitates sites be the dominant recombination centers that exist in the starting material, in which contamination by transition metals may be very low (in the order of 10¹⁰ atoms cm^-3), as may be detected by SPV measurements using the optical separation of the iron-boron pair.

Therefore, lifetime measurements in function of the injection level, in conjunction with lifetime maps confirm the hypothesis that oxygen precipitates represented, at least in the examined samples, the dominant recombination center.

Moreover, the SPV technique is intrinsically suited under conditions of low level injection, and from the confirmatory results obtained from Elymat tests, this technique confirms itself as being extremely sensitive to oxygen precipitation.

According to a standard processing of SPV measurement data, a uniform distribution of the recombination centers is assumed.

In wafers having a superficial denuded zone, an SPV data processing is capable of producing an estimate of the effective diffusion length L_f, resulting from an average between the diffusion length in the denuded zone and the diffusion length in the underlying defective bulk region.

According to an essential aspect of the method of the invention the processing of SPV measurement data on a portion of the surface of a wafer having a denuded zone and an internal or bulk region of where there exist oxygen precipitates and associated defects, the SPV signal data must be processed by considering the sample dishomogeneous in function of depth. This can be done in different ways, for instance through a numerical solution of the equations that govern the evolution of the charge carriers concentration in the semiconductor. There exist computer programs that are specifically suited for this purpose, capable of accounting for the dependence of defects concentration from the depth. The most elementary approach consists in assuming that regions exist wherein:

wherein L_diff is the diffusion length of minority carriers, L_DZ is the diffusion length of minority carriers for a recombination center density of the denuded zone, L_def is the diffusion length of minority carriers in the defective bulk region (L_DZ>>L_def), x is the depth coordinate and t_DZ is the thickness of the denuded zone (DZ).

According to the method of the invention, a direct measure of L_def is obtained by performing the measurements on the backside of the wafer or on the portion of the surface of the wafer where the DZ layer has been purposely stripped off, that is by measuring the diffusion length of minority carrier in the defective bulk region.

In all the tested samples, the measurements taken on the defective bulk region gave lower diffusion lengths than those conducted on the frontside of the wafer, confirming the presence of a relatively defect-free denuded zone on the frontside of the wafer. By contrast, no difference between respective diffusion length measurements was found in wafers with a uniform distribution of recombination centers, for example on samples purposely contaminated with iron.

A quantitative evaluation of the depth of the denuded zone is obtained by solving the diffusion equation of minority charge carriers, while imposing the above stated conditions (1) and continuity of the density function n(x) of the minority carriers and of its first derivative, at the boundary x=tDZ for x equal to t_DZ.

The profiles of the n(x) function are depicted in the diagram of Fig. 4 for the reported testing conditions, namely: t_DZ=40µm, L_def=15µm, penetration depth of incident light from 11 to 157µm.

The L_DZ value is ininfluent as long as the condition L_DZ>>L_def is met and in the calculations the value L_DZ=400 µm was used.

As in a normal SPV technique, the variation of the surface potential signal upon illumination is assumed to be proportional to the concentration of excess minority carriers at the surface, that is correspondent to the function Vph∞n(x=0). This quantity has then been written as a function of the penetration depth z of the particular light being used, with t_DZ, L_def and L_DZ as parameters.

Since, according to the method of the invention, L_def is directly measured and L_DZ is made ininfluent by assuming L_DZ>>L_def, the diffusion function V_ph(z) has only one fit parameter, that is the depth t_DZ of the denuded zone DZ.

By a best-fit procedure of this function, from SPV experimental data, an excellent estimate of the depth (thickness) of the denuded zone is obtained.

Fig. 5 shows the processing results, according to the method of the invention, relative to the case of a sample wafer for which an effective diffusion length of 34µm (L_eff=34µm) was measured on the frontside and a diffusion length of 16µm (L_def=16µm) was measured on the backside of the wafer from where the denuded zone had been previously stripped off. The depth t_DZ of the denuded zone present on the frontside of the wafer calculated according to the method of the invention was 59µm.

As it may be observed from the distribution of the values measured on the wafer frontside (L_eff) and on the wafer backside (L_def), the sample shows a marked dishomogeneity in function of depth from the surface. Despite of this relatively large disuniformity, the curve of the SPV measurements (conventionally /SV in function of 1/α where  is the incident flux and α is the absorption coefficient) appears substantially linear and this was surprisingly observed in all the samples with a denuded zone. This linearity of the curves obtained through SPV measurements at different wavelengths reveals the unreliability of assuming that such a linearity could be indicative of uniformity. The reason for such a scarce correlation between the two aspects may be understood by observing the minority carrier density profiles shown in Fig. 4.

As may be observed, for all the wavelengths of the exciting light, the excess minority carriers profiles extend over a significantly larger depth than that of the denuded zone. Therefore, for all the different penetration depths of the monochromatic lights used, the resulting photovoltage signal derives from an average among the characteristics of density of recombination centers of the denuded layer DZ and of the underlying defective bulk region.

Evidently, it cannot be assumed that the photovoltage signal be dominated by the properties of only one of the two distinct regions.

To fully test the method of the invention, experiments were undertaken on three different sample populations of silicon wafers, subjected and not subjected to a preventive denuding treatment, and for the case of different fabrication processes, naming the three different test sample populations with the letters: a), b) and c), as shown in the following table 1:

LO-HI Process (a) No denuding treatment High temperature diffusion	Process (b) Denuding treatment High temperature diffusion	Process (c) Denuding treatment No further High temp.diffusion
	◆ 1100°C, 4 h, N₂/O₂	◆ 1100°C, 4 h, N₂/O₂
◆ 800°C, 1 h, wet O₂ + 775°C, 1 h 50 min, N₂	◆ 800°C, 1 h, wet O₂ + 775°C, 1 h 50 min, N₂	◆ 800°C, 45 min, wet O₂
◆ 1175°C, 2 h 40 min, N₂/O₂	◆ 1175°C, 2 h 40 min, N₂/O₂	◆ 620°C, 1 h 45 min, N₂
◆ 1100°C; 1 h 40 min, N₂/O₂	◆ 1100°C; 1 h 40 min, N₂/O₂	◆ 775°C, 2 h 30 min, N₂
◆ 800°C, 1 h, wet O₂ + 775°C, h 50 min	◆ 800°C, 45 min, wet O₂	◆ 920°C, 6 h wet O₂ + 1 h N₂
◆ 920°C, 6 h wet O₂ + 1 h N₂	◆ 620°C, 1 h 45 min, N₂
	◆ 775°C, 2 h 15 min, N₂
	◆ 920°C, 6 h wet O₂ + 1 h N₂

The results of the measurements at the edge and at the center of the sample wafers, the semiconductor resistivity, the effective diffusion length (L_def) measured on the wafers backside after stripping off said surface a thickness of material sufficient to ensure a complete removal of any existing denuded layer, the effective diffusion length (L_eff), measured on the frontside of the wafers are shown in table 2, along with the estimated values of the depth of the denuded zone.

ρ (Ωcm)	Process	L_def (µm)	L_eff (µm)	t_dz (µm)
		Edge	Center	Edge	Center	Edge	Center
2	a	15	15.6	28.3	33.3	49.3	58.7
10	a	11.5	8.2	13.1	12.0	15.3	19.9
10	a	13.2	11.2	15.9	13.0	18.9	14.6
10	a	5.15	7.32	18.9	23.0	43.2	49.3
2	b	8.61	7.43	29.4	26.5	61.7	57.1
2	b	26.8	33.6	50.1	51.7	78.4	81.8
2	b	9.67	7.75	29.3	26.5	62.9	56.9
2	b	5.83	6.66	27.5	27.1	61.6	59.7
10	b	21.5	18.8	35.7	37.7	56.6	65.4
10	b	18.4	16.5	34.1	34.6	60.0	62.1
10	b	20.3	18.7	35.9	34.8	59	59.3
2	c	21.7	22.6	25.2	24.1	24.9	17.8
2	c	2.62	2.84	12.7	12.9	32.2	32.5
2	c	21.7	22.5	26.1	23.3	27.7	14.7
2	c	1.69	3.41	11.8	12.9	31.2	31.6
10	c	15.2	16.3	18.0	18.4	20.1	18.1
10	c	12.8	12.8	16.2	15.9	20.9	19.6
10	c	15.9	13.7	17.4	16.5	15.7	19.4

The results obtained were in excellent accordance with microscopic observations that were successively carried out on the sample wafers, thus demonstrating that the method of the invention for estimating the thickness of the denuded zone, is outstandingly reliable.

The nondestructive character of the SPV technique, in conjunction with the relative ease with which the denuded zone may be stripped off, for example, from the entire backside of the wafer by lapping and/or etching, makes the method of the invention suitable for implementing an effective and relatively little costly quality control of the fabrication process, with important beneficial repercussions on the yield.

As already mentioned, the estimation method of the invention, may also make use of measuring techniques of lifetime or effective diffusion length of minority charge carriers different from the SPV techniques, as long as equally effective as the SPV technique, under conditions of a low injection level. For example, the so-called microPCD (microwave-detected Photoconductive Decay ), a technique effective under conditions of low injection may be viable alternative to the SPV technique. There are other techniques similar to micro-PCD techniques, that measure the density of excess minority carriers in function of time based on sample absorption in the infrared .

Claims

Nondestructive method for estimating the depth of a denuded superficial layer in respect to a relatively defective bulk region of a monocrystalline semiconductor wafer by measuring lifetime, that is the effective diffusion length of excess minority charge carriers, under conditions of low injection level, characterized in that comprises the following steps:

stripping off said denuded layer from at least a portion of the surface of the wafer;

separately measuring lifetime or the effective diffusion length of injected excess minority charge carriers on said portion of wafer surface freed of said denuded superficial layer and on a portion of wafer surface having said denuded layer, while injecting said minority charge carriers for a depth sufficient to encompass both said denuded superficial layer and at least a portion of said bulk region;

calculating from the results of said measurements the thickness of said denuded superficial layer by a best-fit procedure or numerical processing.
The method according to claim 1, characterized in that said separate measurements of lifetime or effective diffusion length are carried out with an SPV technique by illuminating an area of the wafer surface with monochromatic light of a wavelength ranging between 800nm and 1000nm.
The method according to claim 2, characterized in that it comprises

expressing the SPV measurement data in function of the penetration depth of the incident monochromatic light and calculating said thickness by a best-fit procedure using as a fitting parameter a thickness value of said denuded superficial layer.